← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9076.a · 生成模型:GPT-6

卷积先乘积后邻域累积

理解等级:中级

实现A与右端时,先形成I_x²、I_xI_y、I_y²、I_xI_t、I_yI_t,再分别与积分窗g卷积。先把各导数独立平滑后再相乘一般不同。对RGB输入,还应明确通道梯度如何合并,不能把标量推导无说明地直接套到三通道张量。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      In order to implement this approach we can make use of convolutions in order to compute all the quantities efficiently in a compact algorithm 48.1: Algorithm 48.1: Gradient-based optical flow estimation using two input frames. 1 Input: ℓ1, ℓ2 ∈ ℝN×M×3, Output: u, v ∈ ℝN×M 2 Parameters: g = integration window 3 Compute: ℓx, ℓy, and ℓt 4 Compute A: 5 Compute b: b1

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第48章《光流估计》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c048.xhtml) - 结论依据:摘录中关于卷积先乘积后邻域累积的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:In order to implement this approach we can make use of convolutions in order to compute all the quantities efficiently in a compact algorithm 48.1: Algorithm 48.1: Gradient-based optical flow estimation using two input frames. 1 Input: ℓ1, ℓ2 ∈ ℝN×M×3, Output: u, v ∈ ℝN×M 2 Parameters: g = integration window 3 Compute: ℓx, ℓy, and ℓt 4 Compute A: 5 Compute b: b1